Short on time? Here’s the whole article in four lines.
- AI can repeat human biases. Read what it gives you with a critical eye, especially about people.
- Copyright around AI is still unsettled, and the rules differ by country. Don’t copy other people’s work, and know that purely AI-made content may not be protected as yours.
- Be honest about how you use AI when someone would reasonably want to know.
- Never use AI to fake proof: no invented reviews, testimonials, or “real” photos.
| Issue | What it means | What you can do |
|---|---|---|
| Bias | AI learned from human writing, so it can repeat stereotypes about gender, race, age, and more | Be specific in your requests, and review anything about people before you use it |
| Copyright | AI was trained on lots of existing work, and the legal rules are still being decided | Don’t ask it to copy a specific artist or writer, and add your own creative work |
| Honesty | People may feel misled if they assume something was made by a person and it wasn’t | Disclose when it matters, and never fake proof |
Ask an AI image tool for “a CEO” or “a nurse,” and you may get very predictable results. Ask for a job description, and the wording may quietly favor one kind of candidate.
Two simple habits help:
Rules vary by country, and some places now require you to label AI content. If you live elsewhere or have customers abroad, here’s a quick snapshot:
| Where | Copyright for AI-made work | Labeling AI content |
|---|---|---|
| European Union | Generally needs a human author’s own creative choices | Under the EU AI Act, from August 2026, realistic AI images, audio, and video (“deepfakes”) must be disclosed, and chatbots must tell people they’re talking to AI. This can apply to you if you have customers in the EU. |
| United Kingdom | An unusual rule can protect “computer-generated” work, treating the person who arranged its creation as the author. The government has been reviewing it. | No general labeling law yet, but advertising rules still forbid misleading people |
| Canada | Unsettled. There’s no clear court ruling yet, and the government has been consulting on it. | No general labeling law yet |
| Australia | Copyright requires a human author | No general labeling law yet |
| China | Courts have protected AI images when a person made significant creative choices | Since September 2025, AI-generated content must be labeled |
The safe habits work everywhere: add your own creative work, label realistic AI images, and check the rules where your customers are.
This is general information, not legal advice.
| Situation | Our suggestion |
|---|---|
| Work a client pays you to create (writing, design) | Talk about it up front. Many clients are fine with it; some contracts require disclosure. |
| Your own social posts and newsletters | Your call. Many people mention it lightly, especially for images. |
| AI images that look like real photos | Label them. |
| Reviews, testimonials, before-and-after results | Never generate these. Use only real ones, with permission. |
| School, grants, contests, publishers | Follow their rules exactly. Many have specific AI policies. |
| Regulated fields (legal, health, financial) | Check your industry’s rules before using AI with clients. |
A good test: would the other person feel misled if they found out? If yes, disclose.
Sam is a freelance copywriter. She uses AI for research and rough first drafts, then rewrites everything in her own voice.
She added one line to her contracts: “I may use AI tools for research and early drafts. All final work is written and reviewed by me.” Not one client objected. Two said it made them trust her more.
Checked October 19, 2026. These rules are changing quickly.
